Effective basis set extrapolations for CCSDT, CCSDT(Q ...
Transcript of Effective basis set extrapolations for CCSDT, CCSDT(Q ...
Effective basis set extrapolations for CCSDT,
CCSDT(Q), and CCSDTQ correlation energies
Amir Karton
School of Molecular Sciences, The University of Western Australia, Perth, WA 6009,
Australia.
ABSTRACT
It is well established that extrapolating the CCSD and (T) correlation energies using
empirically motivated extrapolation exponents can accelerate the basis set convergence. Here
we consider the extrapolation of coupled-cluster expansion terms beyond the CCSD(T) level
to the complete basis set (CBS) limit. We obtain reference CCSDT–CCSD(T) (T3–(T)),
CCSDT(Q)–CCSDT ((Q)), and CCSDTQ–CCSDT(Q) (T4–(Q)) contributions from cc-
pV{5,6}Z extrapolations for a diverse set of 16 first- and second-row systems. We use these
basis-set limit results to fit extrapolation exponents in conjunction with the cc-pV{D,T}Z, cc-
pV{T,Q}Z, and cc-pV{Q,5}Z basis set pairs. The optimal extrapolation exponents result in
noticeable improvements in performance (relative to = 3.0) in conjunction with the cc-
pV{T,Q}Z basis set pair, however, smaller improvements are obtained for the other basis sets.
These results confirm that the basis sets and basis set extrapolations used for obtaining post-
CCSD(T) components in composite thermochemical theories such as Weizmann-4 and HEAT
are sufficiently close to the CBS limit for attaining sub-kJ-per-mole accuracy. The fitted
extrapolation exponents demonstrate that the T3–(T) correlation component converges more
slowly to the CBS limit than the (Q) and T4 terms. A systematic investigation of the effect of
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
2
diffuse functions shows that it diminishes (i) in the order T3–(T) > (Q) > T4–(Q) and (ii) with
the size of the basis set. Importantly, we find that diffuse functions tend to systematically
reduce the T3–(T) contribution but systematically increases the (Q) contribution. Thus, the use
of the cc-pVnZ basis sets benefits from a certain degree of error cancellation between these
two components.
*E-Mail: [email protected]
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
3
1. Introduction
The solution to the nonrelativistic Schrödinger equation can be viewed as a two-
dimensional problem in which both the one-particle space (“basis set completeness”) and n-
particle space (“correlation treatment”) have to be converged to similar levels of accuracy. For
example, in order to obtain thermochemical properties with benchmark accuracy (i.e., with
95% confidence intervals from accurate thermochemical data below ~1 kJ mol–1) both spaces
have to converge to a sub-kJ-per-mole level of accuracy.1,2,3,4,5,6,7,8,9,10 Coupled-cluster (CC)
theory provides a road map for converging the n-particle space,11,12 where the base level is
usually the so-called ‘gold standard’ of computational chemistry – the CCSD(T) method13,14
(i.e., coupled-cluster with single, double, and quasiperturbative triple excitations).2,6,7
However, it is well established that the CCSD(T) method cannot generally achieve benchmark
accuracy for (i) systems dominated by moderate-to-severe non-dynamical correlation effects,
and/or (ii) challenging thermochemical properties such as heats of
formation.5,7,9,10,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33 In such cases post-CCSD(T)
contributions, most importantly up to CCSDT(Q) or CCSDTQ,5,7,9,10,16,20,22,24 have to be added
for achieving high accuracy. The shortcoming of the CCSD(T) method can be illustrated by
examining the magnitude of post-CCSD(T) contributions to the heats of formation in the W4-
17 database (see Table S1 of the supplementary material).16 Of the 200 highly accurate heats
of formation in this representative database, post-CCSD(T) contributions exceed 0.24 kcal
mol–1 (or 1 kJ mol–1) for 50% of the species, 0.5 kcal mol–1 for 24% of the species, and 1 kcal
mol–1 for 7% of the species. Inspections of species with sizeable post-CCSD(T) contributions
reveals that, in addition to multireference systems, they include challenging systems such as
radicals and molecules with highly polar and/or multiple bonds.
The basis set convergence of the CCSD and CCSD(T) correlation energies has been
extensively investigated in conjunction with the correlation-consistent basis sets by Dunning
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
4
and co-workers.2,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56 Owing to the systematic
convergence of these basis sets, basis set extrapolations are an effective approach for
converging the n-particle space while keeping the computational cost at a minimum. However,
the optimal extrapolation parameters may depend on the correlation component being
extrapolated (e.g., CCSD or (T)) and the specific basis sets being used in the extrapolation.48,55
A number of extensive works have obtained optimal extrapolation exponents for the CCSD
and (T) correlation components in conjunction with the correlation-consistent basis
sets,2,34,35,38,46,41,48,49,53,54 where reference data at the infinite basis set limit is obtained from
explicitly correlated (R12 or F12) techniques or extrapolated from very large Gaussian basis
sets.
Less attention has been devoted to obtaining optimal extrapolation exponents for post-
CCSD(T) correlation contributions. Martin and co-workers extrapolated the CCSDT–
CCSD(T) (T3–(T)) correlation component from the cc-pVTZ and cc-pVQZ basis sets for a set
of 15 diatomic molecules and water.22 They found that, relative to cc-pV{Q,5}Z (or where
available cc-pV{5,6}Z) reference data, using an optimized extrapolation exponent of = 2.5
results in a root-mean-square deviation (RMSD) of 0.013 kcal mol–1. This RMSD is nearly half
of that obtained with the asymptotic exponent of = 3.0,57 for which an RMSD of 0.021 kcal
mol–1 is obtained. These results show that extrapolating the T3–(T) correlation component
using effective decay exponents can significantly accelerate the basis set convergence.
Over the past three decades, there has been a proliferation of basis set extrapolation
formulas for the correlation energy. For a historical overview of the various basis set
extrapolations see for example refs. 34, 41, and 43. Here we will primarily focus on the two-
point A + B L–3 expression of Halkier et al.,58 which is motivated by the partial-wave
expansion of pair correlation energies in helium-like atoms57,58,59 and is widely used in
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
5
composite thermochemical protocols, such as the Wn2,7,15,20,23,27,53 Wn-F12,16,42 and
HEAT.21,25,26,60 This leads to the following expression for the extrapolated energy (E):
𝐸∞ = 𝐸(𝐿 + 1) +𝐸(𝐿+1)−𝐸(𝐿)
(𝐿+1
𝐿)
𝛼−1
(1)
where L is the highest angular momentum present in the basis set. We will use this formula
with the asymptotic extrapolation exponent of = 3.0 and with an effective exponent eff which
is optimized to minimize the RMSD for a given coupled-cluster expansion term over our test
set. Schwenke proposed rearranging Eq. 1 to a simplified two-point linear extrapolation which
does not explicitly involve L:
𝐸∞ = 𝐸(𝐿) + 𝐹[𝐸(𝐿 + 1) − 𝐸(𝐿)] (2)
where F is an empirical scaling factor, which is related to via the following expression:
𝐹 = 1 +1
(𝐿+1
𝐿)
𝛼−1
(3)
In the present work we further explore this concept for higher-level correlation components
CCSDT(Q)–CCSDT ((Q)) and CCSDTQ–CCSDT(Q) (T4–(Q)) for which effective
extrapolation coefficients have not been previously obtained. We also re-examine the previous
results22 for the T3–(T) correlation contribution against the larger training set considered in this
work. We begin by calculating the T3–(T), (Q), and T4–(Q) correlation contributions to the
total atomization energies (TAEs) near the one-particle basis set limit (i.e., extrapolated from
the cc-pV{5,6}Z basis set pair) for a representative set of 16 molecules: BH3, CH, CH2(1A1),
CH3, OH, H2O, HF, AlH3, H2S, HCl, B2, C2(1∑+), CO, N2, CS, and P2. Our set of molecules
includes first and second-row hydride and non-hydride systems with 2–4 atoms, including
closed-shell singlet, singlet diradical, radical, and triplet systems. Using these basis set limit
values we obtain effective decay exponent for the T3–(T), (Q), and T4–(Q) correlation
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
6
components in conjunction with the cc-pV{D,T}Z, cc-pV{T,Q}Z, and cc-pV{Q,5}Z basis set
pairs. We also address questions such as:
➢ How does the effective convergence rate vary between the T3–(T), (Q), and ��4
correlation components?
➢ How do the effective decay exponents compare with the asymptotically limiting L–3
convergence behavior?
➢ How does the effective convergence rate differ between hydride and nonhydride
systems?
➢ To what extent does the addition of diffuse functions affect the post-CCSD(T)
contributions?
2. Computational Methods
All calculations were carried out using the MRCC program suite61,62 with the standard
correlation-consistent basis sets of Dunning and co-workers.63,64,65,66 For the sake of brevity,
the cc-pVnZ basis sets (n = D, T, Q, 5, 6) are denoted by VnZ. We also consider the aug'-cc-
pVnZ basis sets, which combine the cc-pVnZ basis sets on H and the aug-cc-pVnZ basis sets
on first- and second-row atoms (denoted by A'VnZ).67 Basis set extrapolations using the VnZ
and V(n+1)Z basis sets are denoted by V{n,n+1}Z. All calculations are nonrelativistic and are
carried out within the frozen-core approximation, i.e., the 1s orbitals for first-row atoms and
the 1s, 2s, and 2p orbitals for second-row atoms are constrained to be doubly occupied in all
configurations. All the geometries were optimized at the CCSD(T)/cc-pV(Q+d)Z level68 of
theory and were taken from the W4-17 database16 (the geometries are given in Table S2 of the
supplementary material). This level of theory has been shown to yield geometries that are in
close agreement with CCSD(T) geometries near the complete basis set (CBS) limit, e.g., with
mean absolute deviations of 0.001 Å from CCSD(T)/cc-pV(6+d)Z geometries.69
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
7
In the present study we examine the basis set convergence of post-CCSD(T)
contributions to molecular atomization energies near the one-particle basis set limit. We
consider iterative and perturbative coupled cluster contributions up to connected quadruples
(CCSDTQ) for a set of 16 first- and second- row molecules with up to four atoms. Table 1 lists
the systems considered here along with the %TAE[(T)] diagnostics for nondynamical
correlation effects.7,10,24 The %TAE[(T)] diagnostics is defined as the percentage of the total
atomization energy accounted for by parenthetical connected triple excitations and has been
found to be a reliable energy-based diagnostic for the importance of non-dynamical correlation
effects.7,10,24 The %TAE[(T)] values suggest that the chosen set of molecules spans the gamut
from systems dominated by dynamical correlation (e.g., H2O and HF), moderate nondynamical
correlation (e.g., N2 and P2), and strong nondynamical correlation (e.g., B2 and C2).7,10,24
Table 1. Overview of the molecules considered in the present work with %TAE[(T)]
diagnostics for nondynamical correlation effects.
Name %TAE[(T)]
BH3 0.3
HC• 1.0
H2C(1A1) 1.0
H3C• 0.6
HO• 1.6
H2O 1.5
HF 1.5
AlH3 0.1
H2S 1.2
HCl 1.4
B2(3Σ𝑔) 14.7
C2(1Σ𝑔+) 13.3
CO 3.1
N2 4.2
CS 5.6
P2 8.3
Table 2 gives an overview of the coupled cluster excitations that are considered in the
present work along with the abbreviations that are used. In particular, we consider the following
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
8
post-CCSD(T) contributions: CCSDT–CCSD(T) (T3–(T)), CCSDT(Q)–CCSDT ((Q)),
CCSDTQ–CCSDT(Q) (T4–(Q)), and CSDTQ–CCSDT (T4).
Table 2. Overview of the post-CCSD(T) contributions discussed in the present work.
Name Definition Abbreviation
Higher-order connected triples CCSDT–CCSD(T) T3–(T)
Noniterative connected quadruples CCSDT(Q)–CCSDT (Q)
Higher-order connected quadruples CCSDTQ–CCSDT(Q) T4–(Q)
Iterative connected quadruples CCSDTQ–CCSDT T4
Finally, we note that the present work represents an extensive computational effort for
obtaining basis set limit values for post-CCSD(T) contributions for systems with 2–4 atoms.
We note that many of the calculations reported here strained our computational resources to
the absolute limit. For example, the fully iterative CCSDTQ/V6Z calculations involve 0.7109
(CH), 2.0109 (B2), 4.7109 (OH), 8.3109 (CH2), 9.3109 (HF and HCl), and 1.01010 (C2)
amplitudes. These calculations ran on dual Intel Xeon machines with up to 1024 GB of RAM
for 3 (CH), 16 (HF), 22 (HCl), 24 (OH), 28 (B2), 42 (CH2), and 46 (C2) days.
3. Results and Discussion
3.1 Reference CBS limit values. Table 3 lists the T3–(T), (Q), and T4–(Q), and T4 CBS
reference values used throughout this work for the parameterization of basis set extrapolation
exponents. The T3–(T) and (Q) contributions are extrapolated to the infinite basis set limit from
the V{5,6}Z basis set pair. For seven systems (CH, CH2, OH, HF, HCl, B2, and C2) we were
able to extrapolate the T4–(Q) contribution from the V{5,6}Z basis set pair. For these seven
systems, the V{Q,5}Z extrapolation results in an RMSD of merely 0.0006 kcal mol–1 relative
to the V{5,6}Z basis set limit values. Where the largest deviation being of 0.0010 kcal mol–1
is obtained for B2. In light of these results, our best T4–(Q) reference values are calculated with
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
9
the V{5,6}Z basis set pair for the above seven systems and with the V{Q,5}Z basis set pair for
the rest of the systems.
Table 3. Complete basis set limit reference values for the T3–(T), (Q), T4–(Q), and T4
contributions to total atomization energies (TAEs) extrapolated from the V{5,6}Z basis set pair
using eq. (1) with an asymptotic extrapolation exponent of = 3.0 (in kcal mol–1).a
Molecule T3–(T) (Q) T4–(Q)a ��4b
BH3 –0.024 0.035 [0.004] 0.038
CH 0.100 0.034 0.003 0.037
CH2(1A1) 0.180 0.117 0.012 0.129
CH3 –0.037 0.064 [–0.001] 0.063
OH –0.035 0.097 –0.007 0.090
H2O –0.243 0.231 [–0.025] 0.206
HF –0.165 0.129 –0.015 0.114
AlH3 –0.016 0.030 [0.006] 0.036
H2S –0.110 0.166 [–0.002] 0.164
HCl –0.141 0.117 –0.002 0.115
B2 0.077 1.253 –0.012 1.242
C2(1∑+) –2.291 3.426 –1.150 2.276
CO –0.581 0.727 [–0.101] 0.626
N2 –0.778 1.197 [–0.176] 1.021
CS –0.663 1.153 [–0.134] 1.019
P2 –0.974 1.649 [–0.189] 1.460 aValues in square brackets are extrapolated from the V{Q,5}Z basis set pair. For the seven systems for which we
have both V{Q,5}Z and V{5,6}Z values the RMSD between the two amounts to merely 0.0006 kcal mol–1 (see
text). bObtained from the values in the (Q) and T4–(Q) columns.
The T3–(T), (Q), and T4–(Q) reference values in Table 3 are extrapolated to the infinite
basis set limit using eq. (1) where in all cases an asymptotic extrapolation exponent of = 3.0
has been used. Apart from the T4–(Q) component for the larger systems, we were able to obtain
the reference values from the V{5,6}Z basis set pair. Nevertheless, an optimal empirical
exponent for this basis set pair is still likely to deviate from an exponent of = 3.0 and possibly
to vary between each of the correlation contributions (T3–(T), (Q), and T4–(Q)). It is therefore
of interest to estimate by how much the V{5,6}Z CBS values change by varying the exponent
from the asymptotic value? A practical approach for examining the effect of using = 3 in the
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
10
V{5,6}Z extrapolations (rather than an optimal exponent), is to use the optimal exponents for
the V{Q,5}Z extrapolations: 2.7342 (T3–(T)), 3.4216 ((Q)), and 3.1381 (T4–(Q)) (Table 5, vide
infra). The effect of varying the exponents for the V{5,6}Z extrapolations for each of the post-
CCSD(T) correlation contributions is presented in Table S3 of the Supporting Information. For
the T3–(T) component changing from 3.0 to 2.734 results in changes in the CBS values
ranging from 0.0003 (CH) to 0.007 (C2) kcal mol–1. These changes represent a fraction of a
percent of the T3–(T)/V{5,6}Z CBS values listed in Table 3. For the (Q) component changing
from 3.0 to 3.422 results in changes in the CBS values ranging from 0.0001 (AlH3) to 0.004
(P2) kcal mol–1. Again, these changes represent a fraction of a percent of the (Q)/V{5,6}Z CBS
values listed in Table 3. The changes for the T4–(Q) component are smaller than 0.001 kcal
mol–1. These small changes in the extrapolated V{5,6}Z CBS values are expected since the
effect of the extrapolation exponent on the extrapolated energy becomes less pronounced as
we approach the infinite basis set limit (or as E(L) → E(L+1) in eqs. (1) and (2)).
It is of interest to examine the magnitude of the basis set limit T3–(T) and T4
contributions (Table 3). It is well known that the T3–(T) contribution tends to universally
decrease the TAEs, whereas the T4 contribution tends to increase them.7,10,15,16,23,24,25 With three
exceptions (B2, C2, and CH2), the T3–(T) contribution decreases the TAEs by amounts ranging
from –0.016 (AlH3) to –2.291 (C2) kcal mol–1. With no exception, the T4 contribution increases
the TAEs by amounts ranging from +0.036 (AlH3) to +2.276 (C2) kcal mol–1. Overall, due to
error cancellation between the T3–(T) and T4 components, post-CCSD(T) contributions tend to
increase the atomization energies by up to 0.486 (P2) kcal mol–1. A striking exception to this is
B2, for which there is no error cancellation between the T3–(T) and T4 components, and as a
result post-CCSD(T) contributions amount to as much as +1.318 kcal mol–1.
Inspection of the CBS values in Table 3 reveals that the magnitude of the T3–(T), (Q),
and T4–(Q) components are systematically larger for non-hydride systems than for hydride
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
11
systems. The magnitude of the T3–(T) component (in absolute value) ranges between 0.016–
0.243 kcal mol–1 for the hydride systems and between 0.077–2.291 kcal mol–1 for the non-
hydride systems. The (Q) component ranges between 0.030–0.231 (hydrides) and 0.727–3.426
(non-hydrides) kcal mol–1. The T4–(Q) component ranges between 0.001–0.025 (hydrides) and
0.012–1.150 (non-hydrides) kcal mol–1.
3.2 Overview of basis set convergence of post-CCSD(T) excitations. Figure 1 depicts the
overall RMSDs for the VnZ basis sets (n = D–6) for the T3–(T), (Q), and T4–(Q) components.
This plot demonstrates the decrease in RMSD for successively higher coupled-cluster
expansion terms. In particular, for any given basis set size (D, T, Q, 5, and 6), the RMSD for
the (Q) component is roughly half of that for the T3–(T) component, and the RMSD for the T4–
(Q) component is roughly one tenth of that for the T3–(T) component.
Figure 1. Basis set truncation errors for the T3–(T), (Q), and T4–(Q) components calculated in
conjunction with the VnZ basis sets (n = D, T, Q, 5, 6). The tabulated values are RMSDs over
the entire set of 16 molecules relative to the basis set limit values in Table 3. The RMSDs are
given in Table 4.
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
12
Table 4 lists the RMSDs for the VnZ basis sets (n = D–6) for the T3–(T), (Q), T4–(Q),
and T4 components. These RMSDs show that the (Q) component converges somewhat faster
than the T3–(T) component. It is instructive to look at the ratios between the RMSDs for
consecutive basis sets, i.e., RMSD[V(n+1)Z]/RMSD[VnZ] (Table 4). For the T3–(T)
component, these ratios are 0.34 (VTZ/VDZ), 0.49 (VQZ/VTZ), 0.55 (V5Z/VQZ), and 0.58
(V6Z/V5Z). For the (Q) component, these ratios are somewhat smaller indicating faster
convergence, namely they are 0.32 (VTZ/VDZ), 0.38 (VQZ/VTZ), 0.47 (V5Z/VQZ), and 0.58
(V6Z/V5Z). Similar conclusions regarding the faster convergence of the (Q) component are
drawn from examining the optimized extrapolation exponents in Table 5 (vide infra). For the
T4–(Q) component the ratio between the RMSDs for consecutive basis sets is nearly constant
at ~0.5 (Table 4). Namely, the RMSD is roughly halved with each increase in the highest
angular momentum present in the basis set.
Table 4. RMSDs (in kcal mol–1) over the entire set of 16 molecules relative to the basis set
limit values in Table 3 for the T3–(T), (Q), T4–(Q), and T4 components calculated in
conjunction with the VnZ basis sets (n = D, T, Q, 5, 6).
Basis set T3–(T) (Q) ��4–(Q) ��4
VDZ 0.490 0.301 0.040 0.265
VTZ 0.168 0.097 0.022 0.077
VQZ 0.083 0.037 0.011 0.027
V5Z 0.045 0.017 0.005 0.012
V6Z 0.026 0.010
RMSD Ratiosa
VTZ/VDZ 0.34 0.32 0.55 0.29
VQZ/VTZ 0.49 0.38 0.50 0.35
V5Z/VQZ 0.55 0.47 0.45 0.44
V6Z/V5Z 0.58 0.58 aRatio between the RMSDs calculated with the V(n+1)Z and VnZ basis sets.
3.3 Effective exponents for higher-order connected triple excitations. Table 5 summarizes
the optimal exponents for the post-CCSD(T) contributions (T3–(T), (Q), T4–(Q), and T4). Table
6 lists deviations for the VnZ basis sets (n = D–6) and V{D,T}Z, V{T,Q}Z, and V{Q,5}Z basis
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
13
set extrapolations from the basis set limit T3–(T)/V{5,6}Z values listed in Table 3. The T3–(T)
component converges smoothly but slowly to the CBS limit. For example, the following
RMSDs are obtained relative to the CBS reference values: 0.490 (VDZ), 0.168 (VTZ), 0.083
(VQZ), 0.045 (V5Z), and 0.026 (V6Z) kcal mol–1 (Table 6). Inspection of the individual errors
for each of the molecules reveals that, with the exception of BH3 and AlH3, all systems exhibit
monotonic basis set convergence. For BH3 and AlH3 the VDZ basis set exhibits anomalous
basis set convergence, which results in large errors for the V{D,T}Z extrapolation. For both
systems, however, the basis set convergence becomes monotonic from the VTZ basis set
onwards.
Table 5. Optimal exponents for the A + B L– extrapolation (eq. 1) and scaling factors (F)
for the Schwenke-type linear extrapolation (eq. 2) for the T3–(T), (Q), T4–(Q), and ��4
contributions.
Basis set T3–(T) (Q) T4–(Q) T4
A + B L– extrapolation (eq. 1)
V{D,T}Z 2.7174 2.9968 1.7139 3.3199
V{T,Q}Z 2.4807 3.3831 2.6072 3.6401
V{Q,5}Z 2.7342 3.4216 3.1381a 4.0408a
Schwenke-type extrapolation (eq. 2)
V{D,T}Z 1.4976 1.4218 1.9964 1.3518
V{T,Q}Z 1.9602 1.6073 1.8952 1.5407
V{Q,5}Z 2.1896 1.8728 1.9860a 1.6832a aObtained for a subset of seven systems for which we have both V{Q,5}Z and V{5,6}Z values.
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
14
Table 6. Convergence of the T3–(T) contribution to the total atomization energy for the set of
16 molecules relative to basis set limit values from V{5,6}Z extrapolations (deviations and
error statistics are given in kcal mol–1).
Basis set VDZ VTZ VQZ V5Z V6Z V{D,T}Z V{D,T}Z V{T,Q}Z V{T,Q}Z V{Q,5}Z V{Q,5}Z
a 3.0000 2.7174c 3.0000 2.4807c 3.0000 2.7342c
Fb 1.4211 1.4976c 1.7297 1.9602c 2.0492 2.1896c
BH3 0.016 0.023 0.010 0.005 0.003 0.025 0.026 0.001 –0.001 0.000 –0.001
CH 0.025 0.012 0.006 0.004 0.002 0.007 0.006 0.002 0.001 0.001 0.001
CH2(1A1) 0.075 0.038 0.020 0.011 0.006 0.022 0.019 0.007 0.003 0.002 0.000
CH3 0.095 0.040 0.018 0.010 0.006 0.017 0.013 0.002 –0.003 0.001 0.000
OH 0.085 0.038 0.019 0.010 0.006 0.018 0.015 0.005 0.000 0.001 –0.001
H2O 0.274 0.108 0.051 0.025 0.014 0.039 0.026 0.010 –0.003 –0.003 –0.007
HF 0.174 0.072 0.033 0.016 0.009 0.029 0.021 0.005 –0.004 –0.002 –0.004
AlH3 0.026 –0.019 –0.011 –0.009 –0.005 –0.038 –0.042 –0.006 –0.004 –0.007 –0.007
H2S 0.206 0.059 0.025 0.017 0.010 –0.003 –0.015 0.001 –0.006 0.007 0.006
HCl 0.163 0.055 0.023 0.014 0.008 0.010 0.002 –0.001 –0.008 0.005 0.003
B2 0.563 0.282 0.140 0.073 0.042 0.164 0.142 0.036 0.004 0.003 –0.007
C2(1∑+) 1.048 0.379 0.185 0.097 0.056 0.097 0.045 0.043 –0.002 0.004 –0.008
CO 0.603 0.193 0.090 0.045 0.026 0.020 –0.011 0.014 –0.009 –0.002 –0.008
N2 0.660 0.195 0.095 0.051 0.030 0.000 –0.036 0.022 –0.001 0.005 –0.001
CS 0.775 0.249 0.121 0.071 0.041 0.028 –0.012 0.028 –0.002 0.018 0.011
P2 0.906 0.252 0.132 0.082 0.048 –0.023 –0.073 0.044 0.016 0.030 0.023
RMSDalld,e 0.490 0.168 0.083 0.045 0.026 0.052 0.046 0.021 0.006 0.010 0.008
MADalld,e 0.356 0.126 0.061 0.034 0.020 0.034 0.032 0.014 0.004 0.006 0.006
MSDalld,e 0.356 0.124 0.060 0.033 0.019 0.026 0.008 0.013 –0.001 0.004 0.000
RMSDhydd,f 0.141 0.054 0.025 0.013 0.008 0.024 0.021 0.005 0.004 0.004 0.004
RMSDnonhydd,g 0.779 0.266 0.131 0.072 0.042 0.079 0.070 0.033 0.008 0.015 0.012
aExtrapolation exponent () used in two-point L– extrapolation eq. (1). bExtrapolation coefficient (F) used in
two-point Schwenke extrapolation eq. (2). cUsing empirical and F parameters optimized to minimize the RMSD. dError statistics with respect to the V{5,6}Z reference values in Table 3. RMSD = root-mean-square deviation,
MAD = mean absolute deviation, MSD = mean signed deviation. eOver all species. fOver hydride species. gOver
non-hydride species.
The two-point extrapolation formula with = 3.0 results in RMSDs of 0.052
(V{D,T}Z), 0.021 (V{T,Q}Z), and 0.010 (V{Q,5}Z) kcal mol–1. Minimizing these RMSDs by
varying the extrapolation exponent results in modest improvements in the RMSDs for the
V{D,T}Z and V{Q,5}Z basis set pairs, namely the RMSDs are reduced by 0.006 and 0.002
kcal mol–1, respectively (Table 6). However, the RMSD for the V{T,Q}Z basis set pair is
reduced from 0.021 (with = 3.0) to merely 0.006 kcal mol–1 (with = 2.4807 or equivalently
F = 1.9602). This empirical extrapolation exponent is in excellent agreement with the empirical
extrapolation exponent of = 2.5 obtained previously using T3–(T)/V{Q,5}Z (and where
available V{5,6}Z) reference data.22 We note that in contrast to the optimized extrapolation
exponents, which do not vary monotonically along the series V{D,T}Z → V{T,Q}Z →
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
15
V{Q,5}Z, the Schwenke the extrapolation parameter clearly reflects the monotonic
convergence of the extrapolation parameters along this series (Table 5).
The bottom two lines of Table 6 give the RMSDs over the subset of 10 hydride
(RMSDhyd) and 6 non-hydride (RMSDnonhyd) systems. Inspection of these RMSDs reveals that
for all the VnZ basis sets (n = D–6), the RMSDs over the hydride systems (RMSDhyd) are
consistently about ~5 times smaller than the RMSDs over the non-hydride systems
(RMSDnonhyd). Thus, the basis set convergence for the hydride systems is significantly faster
than for non-hydride systems. The stark difference between the basis set convergence of the
T3–(T) component for the hydride and non-hydride systems is depicted in Figure 2a.
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
16
Figure 2. Basis set truncation errors for the T3–(T), (Q), and T4–(Q) components calculated in
conjunction with the VnZ basis sets (n = D, T, Q, 5, 6). The tabulated values are RMSDs from
the basis set limit values in Table 3.
3.4 Effective exponents for noniterative connected quadruples excitations. Table 7 lists
deviations for the VnZ basis sets (n = D–6) and V{D,T}Z, V{T,Q}Z, and V{Q,5}Z basis set
extrapolations from the basis set limit (Q)/V{5,6}Z values listed in Table 3. As discussed in
Section 3.2, the (Q) component converges somewhat faster to the CBS limit than the T3–(T)
component. The following RMSDs are obtained relative to the CBS reference values: 0.301
(VDZ), 0.097 (VTZ), 0.037 (VQZ), 0.017 (V5Z), and 0.010 (V6Z) kcal mol–1 (Table 7).
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
17
Table 7. Convergence of the (Q) contribution to the total atomization energy for the set of 16
molecules relative to basis set limit values from V{5,6}Z extrapolations (deviations and error
statistics are given in kcal mol–1).
Basis set VDZ VTZ VQZ V5Z V6Z V{D,T}Z V{D,T}Z V{T,Q}Z V{T,Q}Z V{Q,5}Z V{Q,5}Z
a 3.0000 2.9968c 3.0000 3.3831c 3.0000 3.4216c
Fb 1.4211 1.4218c 1.7297 1.6073c 2.0492 1.8728c
BH3 –
0.013 –0.008 –0.002 –0.001
–
0.001 –0.006 –0.006 0.002 0.001 0.000 0.000
CH –
0.008 –0.007 –0.003 –0.001
–
0.001 –0.007 –0.007 0.000 –0.001 0.000 0.000
CH2(1A1) –
0.035 –0.019 –0.008 –0.004
–
0.002 –0.013 –0.013 0.001 –0.001 0.001 0.000
CH3 –
0.006 –0.015 –0.005 –0.002
–
0.001 –0.019 –0.019 0.002 0.001 0.001 0.000
OH 0.016 –0.019 –0.009 –0.004
–
0.002 –0.034 –0.034 –0.002 –0.004 0.001 0.000
H2O 0.031 –0.039 –0.018 –0.008
–
0.005 –0.069 –0.069 –0.003 –0.005 0.003 0.001
HF 0.060 –0.020 –0.009 –0.004
–
0.002 –0.053 –0.053 –0.002 –0.003 0.001 0.000
AlH3 –
0.014 –0.007 –0.002 –0.001
–
0.001 –0.003 –0.003 0.001 0.000 0.000 0.000
H2S –
0.082 –0.033 –0.016 –0.008
–
0.005 –0.013 –0.013 –0.003 –0.006 0.000 –0.002
HCl –
0.050 –0.025 –0.013 –0.006
–
0.004 –0.014 –0.014 –0.004 –0.005 0.000 –0.001
B2 –
0.345 –0.091 –0.034 –0.014
–
0.008 0.016 0.017 0.008 0.001 0.006 0.003
C2(1∑+) –
0.770 –0.206 –0.074 –0.032
–
0.019 0.032 0.032 0.022 0.006 0.012 0.004
CO –
0.093 –0.075 –0.027 –0.012
–
0.007 –0.067 –0.067 0.009 0.003 0.004 0.001
N2 –
0.169 –0.108 –0.040 –0.019
–
0.011 –0.082 –0.082 0.009 0.001 0.003 –0.001
CS –
0.563 –0.175 –0.071 –0.034
–
0.020 –0.012 –0.012 0.005 –0.008 0.005 –0.001
P2 –
0.609 –0.218 –0.082 –0.041
–
0.024 –0.054 –0.053 0.017 0.001 0.003 –0.005
RMSDalld,e 0.301 0.097 0.037 0.017 0.010 0.040 0.040 0.008 0.004 0.004 0.002
MADalld,e 0.179 0.067 0.026 0.012 0.007 0.031 0.031 0.006 0.003 0.003 0.001
MSDalld,e
–
0.166 –0.067 –0.026 –0.012
–
0.007 –0.025 –0.025 0.004 –0.001 0.003 0.000
RMSDhydd,f 0.040 0.022 0.010 0.005 0.003 0.031 0.031 0.002 0.003 0.001 0.001
RMSDnonhydd,g 0.489 0.156 0.059 0.028 0.016 0.051 0.051 0.013 0.004 0.006 0.003
aExtrapolation exponent () used in two-point L– extrapolation eq. (1). bExtrapolation coefficient (F) used in
two-point Schwenke extrapolation eq. (2). cUsing empirical and F parameters optimized to minimize the RMSD. dError statistics with respect to the V{5,6}Z reference values in Table 3. RMSD = root-mean-square deviation,
MAD = mean absolute deviation, MSD = mean signed deviation. eOver all species. fOver hydride species. gOver
non-hydride species.
The two-point extrapolation formula with = 3.0 results in RMSDs of 0.040
(V{D,T}Z), 0.008 (V{T,Q}Z), and 0.004 (V{Q,5}Z) kcal mol–1. Interestingly, the optimal
exponent for the V{D,T}Z extrapolation ( = 2.9968 or equivalently F = 1.4218) is practically
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
18
identical to the ideal exponent. Minimizing the RMSDs by varying the extrapolation exponents
for the V{T,Q}Z and V{Q,5}Z basis set pairs results in an extrapolation exponent of ~3.4 in
both cases. In particular, we obtain = 3.3831 or equivalently F = 1.6073 for the V{T,Q}Z
extrapolation and = 3.4216 or equivalently F = 1.8728 for the V{Q,5}Z extrapolation. In
both cases, varying the extrapolation exponent from 3.0 to ~3.4 halves the RMSD, and results
in RMSDs of merely 0.004 and 0.002 kcal mol–1 for the V{T,Q}Z and V{Q,5}Z extrapolations,
respectively (Table 7).
In computationally economical post-CCSD(T) methods (e.g., W4lite theory and HEAT-
345(Q))24,25 T4 term is simply calculated as (Q)/VDZ. In W4lite theory this term is scaled by
1.1 to compensate for basis set incompleteness and the complete neglect of the T4–(Q)
component. As mentioned above, the RMSD for the (Q)/VDZ term relative to our (Q)/V{5,6}Z
values is 0.301 kcal mol–1, scaling the (Q)/VDZ term by a scaling factor of 4/3 optimized to
minimize the RMSD results in an RMSD of 0.136 kcal mol–1. Thus, for large systems, for
which the (Q)/VTZ calculation is not feasible, we recommend scaling the (Q)/VDZ term by a
scaling factor of 4/3. In case the (Q)/VDZ term is used for approximating the overall T4/CBS
energy, we find a slightly lower optimal scaling factor of 1.25, which results in an RMSD of
0.129 kcal mol–1 relative to the T4/CBS values. We note that the highly multireference C2
molecule has been excluded from this parameterization.
Inspection of the RMSDs over the subset of 10 hydride (RMSDhyd) and 6 non-hydride
(RMSDnonhyd) systems in Table 7 reveals that the latter are 5–12 times larger than the former.
Thus, similarly to the T3–(T) component, basis set convergence of the (Q) component for the
hydride systems is significantly faster than for non-hydride systems (Figure 2b). However, in
contrast to the T3–(T) component where RMSDnonhyd/RMSDhyd ≈ 5 for all basis sets, for the (Q)
component this ratio is systematically reduced with the size of the basis set. In particular, we
obtain RMSDnonhyd/RMSDhyd = 12.2 (VDZ), 7.1 (VTZ), 5.9 (VQZ), 5.6 (V5Z), and 5.3 (V6Z).
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
19
3.5 Effective exponents for higher-order connected quadruple excitations. Table 5
summarizes the optimal exponents for the ��4–(Q) component. Table 8 lists deviations for the
VnZ basis sets (n = D–5) and the V{D,T}Z and V{T,Q}Z basis set extrapolations from the
basis set limit ��4–(Q)/V{5,6}Z (or ��4–(Q)/V{Q,5}Z) values listed in Table 3. As briefly
discussed in Section 3.3, the T4–(Q) component converges smoothly to the basis set limit with
a reduction of the RMSD by 50% with each increase in the highest angular momentum present
in the basis set. In particular, we obtain the following RMSDs: 0.040 (VDZ), 0.022 (VTZ),
0.011 (VQZ), and 0.005 (V5Z) kcal mol–1 (Table 8). The two-point extrapolation formula with
= 3.0 results in RMSDs of 0.016 (V{D,T}Z) and 0.003 (V{T,Q}Z) kcal mol–1. Minimizing
these RMSDs by varying the extrapolation exponent results in relatively small improvements
in the RMSDs, namely the RMSDs are reduced to 0.011 (V{D,T}Z) and 0.002 (V{T,Q}Z)
(Table 8).
Table 8. Convergence of the T4–(Q) contribution to the total atomization energy for the set of
16 molecules relative to basis set limit values from V{5,6}Z or V{Q,5}Z extrapolations (see
Table 3) (deviations and error statistics are given in kcal mol–1).
Basis set VDZ VTZ VQZ V5Z V{D,T}Z V{D,T}Z V{T,Q}Z V{T,Q}Z
a 3.0000 1.7139 3.0000 2.6072
Fb 1.4211 1.9964 1.7297 1.8952
BH3 0.001 0.001 0.000 0.000 0.001 0.001 0.000 0.000
CH 0.000 0.001 0.000 0.000 0.001 0.001 0.000 0.000
CH2(1A1) 0.002 0.003 0.002 0.001 0.004 0.005 0.001 0.001
CH3 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000
OH 0.007 0.002 0.002 0.001 –0.001 –0.004 0.002 0.002
H2O 0.000 0.003 0.003 0.001 0.004 0.006 0.002 0.002
HF –0.001 –0.001 0.001 0.001 –0.001 –0.001 0.003 0.003
AlH3 0.000 0.000 0.001 0.000 0.001 0.001 0.001 0.001
H2S 0.002 0.006 0.003 0.001 0.008 0.011 0.000 0.000
HCl 0.000 0.004 0.002 0.001 0.005 0.007 0.002 0.001
B2 0.096 0.043 0.021 0.010 0.021 –0.010 0.004 0.001
C2(1∑+) 0.083 0.048 0.022 0.011 0.033 0.013 0.003 –0.001
CO 0.000 0.006 0.003 0.002 0.008 0.012 0.001 0.001
N2 0.019 0.025 0.010 0.005 0.028 0.031 –0.001 –0.004
CS 0.046 0.031 0.017 0.009 0.025 0.017 0.007 0.004
P2 0.082 0.044 0.021 0.011 0.028 0.006 0.004 0.000
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
20
RMSDalld,e 0.040 0.022 0.011 0.005 0.016 0.011 0.003 0.002
MADalld,e 0.021 0.014 0.007 0.004 0.011 0.008 0.002 0.001
MSDalld,e 0.021 0.013 0.007 0.004 0.010 0.006 0.002 0.001
RMSDhydd,f 0.002 0.003 0.002 0.001 0.004 0.005 0.002 0.002
RMSDnonhydd,g 0.065 0.036 0.017 0.009 0.025 0.017 0.004 0.002
aExtrapolation exponent () used in two-point L– extrapolation eq. (1). bExtrapolation coefficient (F) used in
two-point Schwenke extrapolation eq. (2). cUsing empirical and F parameters optimized to minimize the RMSD. dError statistics with respect to the V{5,6}Z reference values in Table 3. RMSD = root-mean-square deviation,
MAD = mean absolute deviation, MSD = mean signed deviation. eOver all species. fOver hydride species. gOver
non-hydride species.
The magnitude of the ��4–(Q) component for hydrides (in absolute value) ranges
between 0.001 (CH3) and 0.025 (H2O) kcal mol–1 and accordingly the RMSDs for the various
basis sets are negligibly small (Table 8). The RMSDs over the subset of 6 non-hydride systems
are significantly larger as illustrated in Figure 2c, specifically RMSDnonhyd = 0.065 (VDZ),
0.036 (VTZ), 0.017 (VQZ), and 0.009 (V5Z) kcal mol–1.
3.6 Effective exponents for iterative connected quadruple excitations. Let us now consider
the basis set convergence of the ��4 component as a whole. These results are summarized in
Table 9. Overall, the basis set convergence of the ��4 component is similar to that of the (Q)
component. In particular, we obtain RMSDs of 0.265 (VDZ), 0.077 (VTZ), 0.027 (VQZ), and
0.012 (V5Z). Thus, the smallest basis set that achieves sub-kJ/mol accuracy is the VTZ basis
set. The two-point extrapolation formula with = 3.0 results in RMSDs of 0.037 (V{D,T}Z)
and 0.010 (V{T,Q}Z) kcal mol–1. Similarly to the (Q) component, optimizing the exponent for
the V{D,T}Z extrapolation results in marginal improvement in performance and leads to an
RMSD of 0.034 kcal mol–1. However, optimizing the exponent for the V{T,Q}Z extrapolation
reduces the RMSD from 0.010 to 0.003 kcal mol–1.
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
21
Table 9. Convergence of the T4 contribution to the total atomization energy for the set of 16
molecules relative to basis set limit values from V{5,6}Z or V{Q,5}Z extrapolations (see Table
3) (deviations and error statistics are given in kcal mol–1).
Basis set VDZ VTZ VQZ V5Z V{D,T}Z V{D,T}Z V{T,Q}Z V{T,Q}Z
a 3.0 3.3199 3.0 3.6401
Fb 1.4211 1.3518 1.7297 1.5407
BH3 –0.012 –0.008 –0.002 –0.001 –0.006 –0.006 0.002 0.001
CH –0.008 –0.006 –0.003 –0.001 –0.006 –0.006 0.000 –0.001
CH2(1A1) –0.033 –0.016 –0.006 –0.002 –0.009 –0.010 0.002 0.000
CH3 –0.006 –0.015 –0.005 –0.002 –0.018 –0.018 0.002 0.001
OH 0.023 –0.017 –0.008 –0.003 –0.035 –0.032 –0.001 –0.003
H2O 0.031 –0.036 –0.016 –0.007 –0.065 –0.060 0.000 –0.004
HF 0.059 –0.021 –0.008 –0.003 –0.054 –0.049 0.001 –0.001
AlH3 –0.014 –0.006 –0.001 –0.001 –0.003 –0.003 0.002 0.001
H2S –0.080 –0.027 –0.013 –0.007 –0.005 –0.009 –0.003 –0.006
HCl –0.050 –0.021 –0.010 –0.005 –0.009 –0.011 –0.002 –0.004
B2 –0.249 –0.048 –0.013 –0.004 0.037 0.023 0.013 0.006
C2(1∑+) –0.687 –0.158 –0.052 –0.021 0.065 0.028 0.026 0.006
CO –0.093 –0.069 –0.023 –0.010 –0.059 –0.061 0.010 0.001
N2 –0.150 –0.083 –0.030 –0.014 –0.054 –0.059 0.008 –0.002
CS –0.517 –0.144 –0.054 –0.025 0.013 –0.013 0.011 –0.006
P2 –0.527 –0.175 –0.061 –0.030 –0.026 –0.051 0.021 0.000
RMSDalld,e 0.265 0.077 0.027 0.012 0.037 0.034 0.010 0.003
MADalld,e 0.159 0.053 0.019 0.008 0.029 0.027 0.007 0.003
MSDalld,e –0.145 –0.053 –0.019 –0.008 –0.015 –0.021 0.006 –0.001
RMSDhydd,f 0.039 0.020 0.008 0.004 0.030 0.028 0.002 0.003
RMSDnonhydd,g 0.430 0.123 0.043 0.020 0.046 0.043 0.016 0.004
aExtrapolation exponent () used in two-point L– extrapolation eq. (1). bExtrapolation coefficient (F) used in
two-point Schwenke extrapolation eq. (2). cUsing empirical and F parameters optimized to minimize the RMSD. dError statistics with respect to the V{5,6}Z reference values in Table 3. RMSD = root-mean-square deviation,
MAD = mean absolute deviation, MSD = mean signed deviation. eOver all species. fOver hydride species. gOver
non-hydride species.
3.7 The effect of diffuse functions on post-CCSD(T) contributions. The steep computational
cost of post-CCSD(T) calculations (particularly CCSDT(Q) and beyond) makes the use of the
aug-cc-pVnZ basis sets unfeasible for systems with more than a handful of atoms. Therefore,
post-CCSD(T) composite theories such as Weizmann-4 and HEAT use augmented basis sets
for the CCSD(T) calculations and regular cc-pVnZ basis sets for the post-CCSD(T)
contributions. Nevertheless, it is of interest to examine the effect of diffuse functions on the
post-CCSD(T) contributions. Here we will consider the T3–(T), (Q), and T4–(Q) components
with basis sets of up to A'V6Z, A'V6Z, and A'VQZ quality, respectively. In addition, it is
important to estimate to what extent the addition of diffuse function affects the CBS values
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
22
reported in Table 3. For this purpose, we consider a subset of nine systems containing
electronegative atoms (N, O, F, P, S, and Cl). Table 10 gives an overview of the effects of
adding diffuse functions on the post-CCSD(T) contributions. The tabulated values are the
differences between the post-CCSD(T) contribution calculated with the VnZ and A'VnZ basis
sets, i.e., VnZ – A'VnZ. Two key points that can be drawn from Table 10 are that the effect of
adding diffuse functions tends to diminish (i) in the order T3–(T) > (Q) > T4–(Q) and (ii) with
each increase in the highest angular momentum represented in the basis set. For any given basis
set, the RMSD for the (Q) contribution is smaller by ~50% than that for the T3–(T) contribution.
Similarly, the RMSD for the T4–(Q) contribution is smaller by about one order of magnitude
than that for the (Q) contribution. For any given correlation component (T3–(T), (Q), and T4–
(Q)), the RMSD is reduced by ~50% with each increase in the highest angular momentum
present in the basis set. The RMSDs for the largest basis set considered are 0.006 (T3–
(T)/A'V6Z), 0.002 ((Q)/A'V6Z), and 0.003 (T4–(Q)/A'VQZ) kcal mol–1. These RMSDs
indicate that the CBS reference values reported in Table 3 should be little affected by the
addition of diffuse function.
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
23
Table 10. Effect of diffuse functions on post-CCSD(T) contributions. The tabulated values are
the differences between the post-CCSD(T) contribution to the total atomization energy
calculated with the VnZ and A'VnZ basis sets, i.e., TAE(VnZ) – TAE(A'VnZ) (in kcal mol–1).
Basis set OH H2O HF H2S HCl CO N2 CS P2 RMSDa
T3–(T) A'VDZ 0.005 0.065 0.028 0.047 0.033 0.165 0.167 0.213 0.167 0.123
A'VTZ 0.008 0.045 0.026 0.018 0.017 0.076 0.079 0.094 0.096 0.061
A'VQZ 0.002 0.017 0.011 0.006 0.005 0.025 0.026 0.032 0.033 0.021
A'V5Z 0.001 0.005 0.004 0.004 0.004 0.009 0.012 0.020 0.023 0.012
A'V6Z 0.001 0.002 0.002 0.002 0.002 0.004 0.006 0.010 0.014 0.006
(Q) A'VDZ –0.005 –0.021 0.004 –0.014 –0.010 –0.049 –0.107 –0.129 –0.100 0.068
A'VTZ –0.005 –0.012 0.000 –0.009 –0.005 –0.016 –0.046 –0.044 –0.045 0.027
A'VQZ –0.004 –0.008 –0.003 –0.004 –0.003 –0.005 –0.019 –0.017 –0.018 0.011
A'V5Z –0.002 –0.003 –0.001 –0.002 –0.002 –0.002 –0.008 –0.007 –0.008 0.005
A'V6Z –0.001 –0.001 0.000 N/A 0.000 N/A –0.003 N/A –0.004 0.002
T4–(Q) A'VDZ 0.003 0.008 0.002 0.004 0.002 –0.006 0.012 0.014 0.023 0.011
A'VTZ 0.001 0.002 –0.001 0.001 0.001 –0.004 0.007 0.007 0.007 0.004
A'VQZ 0.001 0.002 0.000 0.001 0.000 –0.001 0.003 0.005 0.004 0.003
(Q)–(T) A'VDZ 0.000 0.043 0.032 0.033 0.023 0.116 0.060 0.084 0.067 0.061
A'VTZ 0.003 0.033 0.027 0.009 0.012 0.060 0.033 0.050 0.051 0.036
A'VQZ –0.002 0.009 0.008 0.001 0.002 0.020 0.007 0.015 0.015 0.011
A'V5Z –0.001 0.002 0.002 0.002 0.002 0.007 0.004 0.013 0.015 0.007 aRMSD over the TAE(VnZ) – TAE(A'VnZ) differences for the nine molecules.
Another important observation is that the addition of diffuse functions tends to
systematically reduces the T3–(T) contribution but systematically increases the (Q)
contribution. This is a significant result since it indicates that the use of the regular VnZ basis
sets for the calculation of the CCSDT(Q)–CCSD(T) contribution should benefit from a certain
degree of error cancellation between the T3–(T) and (Q) components (see e.g., RMSDs in Table
10).
3.8 Transferability of the optimized exponents. Finally, a comment is due on the
transferability of the optimized exponents in Table 5 to systems outside the training set. In
general, the transferability of the optimal exponents hinges on the selected training set being
reasonably wide and diverse. For example, the selected set of 16 molecules includes first-row,
second-row, hydride, non-hydride, closed-shell singlet, singlet diradical, radical, and triplet
systems. Nevertheless, it is important to test the extendibility of the extrapolation parameters
in Table 5 to systems outside the training set. One way to examine this is by removing six of
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
24
the 16 systems from the training set, reoptimizing the extrapolation exponents, and using the
omitted systems as a test of the extendibility of the extrapolation parameters. We note that the
reduced training set of ten systems is still reasonably diverse and includes BH3, CH3, OH, H2O,
AlH3, H2S, B2, C2, N2, and P2. The RMSDs over the omitted systems using the exponents
optimized over the reduced and entire training sets are compared in Table S4 of the Supporting
Information. Using the exponents optimized over the reduced training set affects the overall
RMSDs by less than 0.001 kcal mol–1. Therefore, we conclude that the extrapolation
parameters in Table 5 should be extendible to similar systems outside our training set of 16
systems.
4. Conclusions
The CCSDT–CCSD(T) (T3–(T)), CCSDT(Q)–CCSDT ((Q)), CCSDTQ–CCSDT(Q)
(T4–(Q)), and CCSDTQ–CCSDT (T4) correlation components are extrapolated to the one-
particle basis set limit from the cc-pV{5,6}Z basis set pair for a diverse set of 16 molecules.
The selected set of molecules includes first-row, second-row, hydride, non-hydride, closed-
shell singlet, singlet diradical, radical, and triplet systems. Using these basis set limit values
we obtain effective decay exponent for the T3–(T), (Q), T4–(Q), and T4 correlation components
in conjunction with the cc-pV{D,T}Z, cc-pV{T,Q}Z, and cc-pV{Q,5}Z basis set pairs. The
following major conclusions can be drawn from the present study:
➢ The fitted extrapolation exponents demonstrate that the T3–(T) correlation component
converges more slowly to the infinite basis set limit than the (Q) and T4 terms.
➢ The optimal extrapolation exponents result in significant improvements in performance
(relative to = 3.0) for the T3–(T), (Q), and T4 components in conjunction with the cc-
pV{T,Q}Z basis set pair. However, smaller improvements in performance are obtained
with the other basis set pairs.
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
25
➢ For the T3–(T) component convergence is slower than the asymptotically limiting L–3
convergence behavior, with effective exponents eff = 2.7174 (cc-pV{D,T}Z), 2.4807
(cc-pV{T,Q}Z), and 2.7342 (cc-pV{Q,5}Z).
➢ The (Q) component tends to converge faster than the L–3 convergence behavior as the
size of the basis sets increases, with effective exponents of eff = 2.9968 (cc-
pV{D,T}Z), 3.3831 (cc-pV{T,Q}Z), and 3.4216 (cc-pV{Q,5}Z).
➢ In cases where only the (Q)/cc-pVDZ calculation is feasible, scaling this component by
1.25 results in an RMSD of 0.13 kcal mol–1 relative to our best T4/CBS values.
➢ The T4 component converges faster than the L–3 convergence behavior and the
convergence rate increases with the size of the basis sets, with effective exponents of
eff = 3.3199 (cc-pV{D,T}Z), 3.6401 (cc-pV{T,Q}Z), and 4.0408 (cc-pV{Q,5}Z).
➢ The effect of adding diffuse functions tends to diminish (i) in the order T3–(T) > (Q) >
T4–(Q) and (ii) with the highest angular momentum represented in the basis set.
➢ The addition of diffuse functions tends to systematically reduces the T3–(T)
contribution and systematically increases the (Q) contribution. Thus, the use of the
regular cc-pVnZ basis sets benefits from a certain degree of error cancellation between
these two components.
➢ Taken together, these results confirm that the basis sets and basis set extrapolations
used for obtaining post-CCSD(T) components in composite thermochemical theories
such as Weizmann-4 and HEAT are sufficiently converged to the complete basis set
limit for attaining sub-kJ-per-mole accuracy.
➢ Across all post-CCSD(T) contributions, the basis set convergence for the hydride
systems is significantly faster than for non-hydride systems.
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
26
Supplementary Material
See supplementary material for post-CCSD(T) contributions for the 200 species in the W4-17
database (Table S1); optimized geometries for all the species considered in this work (Table
S2); effect of varying the extrapolation exponent of = 3.0 for obtaining the CBS limit
reference values in Table 3 (Table S3); examination of the transferability of the optimized
exponents reported in Table 5 (Table S4); CCSDT–CCSD(T) contributions to the total
atomization energies (Table S5); CCSDT(Q)–CCSDT contributions to the total atomization
energies (Table S6); CCSDTQ–CCSDT(Q) contributions to the total atomization energies
(Table S7); and CCSD(T), CCSDT, CCSDT(Q), and CCSDTQ absolute energies calculated in
this work (Table S8).
Acknowledgments
This research was undertaken with the assistance of resources from the National Computational
Infrastructure (NCI), which is supported by the Australian Government. We also acknowledge
system administration support provided by the Faculty of Science at the University of Western
Australia to the Linux cluster of the Karton group. I would like to thank the reviewers of this
manuscript for their constructive comments and suggestions. I gratefully acknowledge an
Australian Research Council (ARC) Future Fellowship (FT170100373).
Data Availability Statement
The data that supports the findings of this study are available within the article [and its
supplementary material]. Any additional data are available from the corresponding author upon
reasonable request.
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
27
References
1 T. Helgaker, W. Klopper, K. L. Bak, A. Halkier, P. Jørgensen, and J. Olsen, “Highly accurate
ab initio computation of thermochemical data,” in Quantum-Mechanical Prediction of
Thermochemical Data, Understanding Chemical Reactivity, edited by J. Cioslowski (Kluwer,
Dordrecht, 2001), Vol. 22, pp. 1–30.
2 J. M. L. Martin and S. Parthiban, “W1 and W2 theory and their variants: Thermochemistry
in the kJ/mol accuracy range,” in Quantum-mechanical Prediction of Thermochemical Data,
Understanding Chemical Reactivity, edited by J. Cioslowski (Kluwer, Dordrecht, 2001), Vol.
22, pp. 31–65.
3 J. M. L. Martin, Annu. Rep. Comput. Chem. 1, 31 (2005).
4 T. Helgaker, W. Klopper, and D. P. Tew, Mol. Phys. 106, 2107 (2008).
5 D. Feller, K. A. Peterson, and D. A. Dixon, J. Chem. Phys. 129, 204105 (2008).
6N. DeYonker, T. R. Cundari, and A. K. Wilson, in Advances in the Theory of Atomic and
Molecular Systems, Progress in Theoretical Chemistry and Physics, edited by P. Piecuch, J.
Maruani, G. Delgado-Barrio, and S. Wilson (Springer Netherlands, Dordrecht, 2009), Vol.
19, pp. 197–224.
7 A. Karton, S. Daon, and J. M. L. Martin, Chem. Phys. Lett. 510, 165 (2011).
8 L. A. Curtiss, P. C. Redfern, and K. Raghavachari, WIREs Comput. Mol. Sci. 1, 810
(2011).
9 K. A. Peterson, D. Feller, and D. A. Dixon, Theor. Chem. Acc. 131, 1079 (2012).
10 A. Karton, WIREs Comput. Mol. Sci. 6, 292 (2016).
11 I. Shavitt and R. J. Bartlett, in Many-Body Methods in Chemistry and Physics: MBPT and
Coupled-Cluster Theory, Cambridge Molecular Science Series (Cambridge University Press,
Cambridge, 2009).
12 R. J. Bartlett and M. Musial, Rev. Mod. Phys. 79, 291 (2007).
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
28
13 K. Raghavachari, G. W. Trucks, J. A. Pople, and M. Head-Gordon, Chem. Phys. Lett. 157,
479 (1989).
14 J. D. Watts, J. Gauss, and R. J. Bartlett, J. Chem. Phys. 98, 8718 (1993).
15 A. Karton, J. Chem. Phys. 149, 034102 (2018).
16 A. Karton, N. Sylvetsky, and J. M. L. Martin, J. Comput. Chem. 38, 2063 (2017).
17 N. Sylvetsky, K. A. Peterson, A. Karton, and J. M. L. Martin, J. Chem. Phys. 144, 214101
(2016).
18D. Feller, K. A. Peterson, and B. Ruscic, Theor. Chem. Acc. 133, 1407 (2014).
19 A. Karton, S. Parthiban, and J. M. L. Martin, J. Phys. Chem. A 113, 4802 (2009).
20 A. Karton, I. Kaminker, and J. M. L. Martin, J. Phys. Chem. A 113, 7610 (2009).
21 M. E. Harding, J. Vazquez, B. Ruscic, A. K. Wilson, J. Gauss, and J. F. Stanton, J. Chem.
Phys. 128, 114111 (2008).
22 A. Karton, P. R. Taylor, and J. M. L. Martin. J. Chem. Phys. 127, 064104 (2007).
23 D. Feller and K. A. Peterson, J. Chem. Phys. 126, 114105 (2007).
24A. Karton, E. Rabinovich, J. M. L. Martin, and B. Ruscic, J. Chem. Phys. 125, 144108 (2006).
25Y. J. Bomble, J. Vazquez, M. Kallay, C. Michauk, P. G. Szalay, A. G. Csaszar, J. Gauss, and
J. F. Stanton, J. Chem. Phys. 125, 064108 (2006).
26A. Tajti, P. G. Szalay, A. G. Csaszar, M. Kallay, J. Gauss, E. F. Valeev, B. A. Flowers, J.
Vazquez, and J. F. Stanton, J. Chem. Phys. 121, 11599 (2004).
27A. D. Boese, M. Oren, O. Atasoylu, J. M. L. Martin, and M. Kallay, J. Gauss, J. Chem. Phys.
120, 4129 (2004).
28T. A. Ruden, T. Helgaker, P. Jørgensen, and J. Olsen, Chem. Phys. Lett. 371, 62 (2003).
29 A. Karton, P. R. Schreiner, and J. M. L. Martin, J. Comput. Chem. 37, 49 (2016).
30 R. O. Ramabhadran and K. Raghavachari, Acc. Chem. Res. 47, 3596 (2014).
31 R. O. Ramabhadran and K. Raghavachari, J. Phys. Chem. A 116, 7531 (2012).
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
29
32 S. E. Wheeler, WIREs Comput. Mol. Sci. 2, 204 (2012).
33 S. E. Wheeler, K. N. Houk, P. V. R. Schleyer, and W. D. Allen, J. Am. Chem. Soc. 131,
2547 (2009).
34 J. M. L. Martin, AIP Conf. Proc. 2040, 020008 (2018).
35 A. J. C. Varandas, Annu. Rev. Phys. Chem. 69, 177 (2018).
36 F. N. N. Pansini, A. C. Neto, and A. J. C. Varandas, Chem. Phys. Lett. 641, 90 (2015).
37 F. N. N. Pansini and A. J. C. Varandas, Chem. Phys. Lett. 631-632, 70 (2015).
38 A. J. C. Varandas and F. N. N. Pansini, J. Chem. Phys. 141, 224113 (2014).
39 D. S. Ranasinghe and E. C. Barnes, J. Chem. Phys. 140, 184116 (2014).
40 D. S. Ranasinghe and G. A. Petersson, J. Chem. Phys. 138, 144104 (2013).
41 D. Feller, J. Chem. Phys. 138, 074103 (2013).
42 A. Karton and J. M. L. Martin, J. Chem. Phys. 136, 124114 (2012).
43 D. Feller, K. A. Peterson, J. G. Hill, J. Chem. Phys. 135, 044102 (2011).
44 J. G. Hill, K. A. Peterson, G. Knizia, and H.-J. Werner, J. Chem. Phys. 131, 194105 (2009).
45 D. Bakowies, J. Chem. Phys. 127, 164109 (2007).
46 D. Bakowies, J. Chem. Phys. 127, 084105 (2007).
47 D. Feller, K. A. Peterson, and T. D. Crawford, J. Chem. Phys. 124, 054107 (2006).
48 D. W. Schwenke, J. Chem. Phys. 122, 014107 (2005).
49 E. F. Valeev, W. D. Allen, R. Hernandez, C. D. Sherrill, and H. F. Schaefer, J. Chem.
Phys. 118, 8594 (2003).
50 D. Feller and D. A. Dixon, J. Chem. Phys. 115, 3484 (2001).
51 W. Klopper, Mol. Phys. 99, 481 (2001).
52 K. L. Bak, P. Jørgensen, J. Olsen, T. Helgaker, and W. Klopper, J. Chem. Phys. 112, 9229
(2000).
53 J. M. L. Martin and G. de Oliveira, J. Chem. Phys. 111, 1843 (1999).
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674
30
54 D. G. Truhlar, Chem. Phys. Lett. 294, 45 (1998).
55 A. G. Császár, W. D. Allen, and H. F. Schaefer III, J. Chem. Phys. 108, 9751 (1998).
56 W. Klopper, J. Noga, H. Koch, and T. Helgaker, Theor. Chem. Acc. 97, 164 (1997).
57 W. Kutzelnigg and J. D. Morgan III, J. Chem. Phys. 96, 4484 (1992); J. Chem. Phys. 97,
8821(E) (1992).
58 A. Halkier, T. Helgaker, P. Jørgensen, W. Klopper, H. Koch, J. Olsen, and A. K. Wilson,
Chem. Phys. Lett. 286, 243 (1998).
59 R. N. Hill, J. Chem. Phys. 83, 1173 (1985).
60 A. Ganyecz, M. Kallay, and J. Csontos, J. Chem. Theory Comput. 13, 4193 (2017).
61 MRCC, a quantum chemical program suite written by M. Kallay, et al., See also:
http://www.mrcc.hu.
62 Z. Rolik, L. Szegedy, I. Ladjanszki, B. Ladoczki, and M. Kallay, J. Chem. Phys. 139, 094105
(2013).
63 T. H. Dunning, J. Chem. Phys. 90, 1007 (1989).
64 R. A. Kendall, T. H. Dunning, Jr., and R. J. Harrison, J. Chem. Phys. 96, 6796 (1992).
65 A. K. Wilson, T. van Mourik, and T. H. Dunning, J. Mol. Struct.: THEOCHEM 388, 339
(1996).
66 T. van Mourik and T. H. Dunning, Int. J. Quantum Chem. 76, 205 (2000).
67 R. A. Kendall, T. H. Dunning, and R. J. Harrison, J. Chem. Phys. 96, 6796 (1992).
68 T. H. Dunning, K. A. Peterson, and A. K. Wilson, J. Chem. Phys. 114, 9244 (2001).
69 P. R. Spackman, D. Jayatilaka, and A. Karton, J. Chem. Phys. 145, 104101 (2016).
Th
is is
the au
thor’s
peer
revie
wed,
acce
pted m
anus
cript.
How
ever
, the o
nline
versi
on of
reco
rd w
ill be
diffe
rent
from
this v
ersio
n onc
e it h
as be
en co
pyed
ited a
nd ty
pese
t. PL
EASE
CIT
E TH
IS A
RTIC
LE A
S DO
I: 10.1
063/5
.0011
674